FRBSF Economic Letter
2003-35; November 28, 2003
Monitoring Debt Market Information for Bank Supervisory Purposes
Bank supervisors monitor bank holding companies (BHCs) in order
to enforce regulations and gauge their soundness so as to guard against
systemic
risk in the financial system. This monitoring is chiefly conducted using
supervisory resources, such as bank examinations and quarterly filings
of balance sheet information. BHCs are also monitored by other parties,
in particular investors in their public securities; this type of monitoring
is commonly referred to as market discipline, and the market assessments
from it are reflected in the prices of BHC public securities.
In the past few years, supervisors have expressed an intent to incorporate
more market information into their monitoring efforts. Researchers and
policymakers have advocated using BHC subordinated debt as a tool for
increasing the degree of market discipline applied to BHCs. A recent
study by the Board of Governors of the Federal Reserve System and the
Secretary of the Treasury (2000) found evidence supporting this assertion
and concluded that subordinated debt issuance might enhance both direct
and indirect market discipline.
In this Economic Letter, we present empirical evidence on the potential
usefulness of BHC debt market information, and securities market information
more broadly, for supervisory monitoring. We find that changes in BHC
bond yields precede corresponding changes in supervisory BHC ratings
by a year. We also assess the contribution of BHC debt yields in the
context of an off-site monitoring model that includes both supervisory
and equity market variables. Our results indicate that debt market information
is useful for supervisory monitoring, even in the presence of equity
market data. In fact, debt yields contributed more explanatory power
for BHCs that are relatively closer to insolvency.
Debt market information
and supervisory monitoring
Publicly traded BHC debt, and subordinated
debt in particular, may be useful for supervisory monitoring because,
in general, BHC debtholders
have incentives roughly in line with supervisors. That is, investors
in BHC subordinated debt are among the first to lose their investments,
after equity investors, if the bank defaults, but they do not benefit
directly from any gains that accrue beyond the face value of their holdings.
Like supervisors, these investors have an incentive to monitor BHC conditions,
and their assessments could be imputed from changes in public BHC debt
prices. However, as a BHC nears insolvency, the incentives of subordinated
debt holders become more like those of equity investors, who are willing
to participate in riskier transactions to "save" the concern.
Such transactions would not be aligned with the concerns of supervisors,
who would be most interested in safeguarding the existing assets. Note,
however, that changes in equity prices still should contain information
on investor assessments of BHC performance that could be useful for supervisory
monitoring.
We gauged the potential supervisory value of BHC debt market information
by incorporating it into models of BHC supervisory ratings. The Federal
Reserve supervises U.S. BHCs, which make up the bulk of public bond issuance
by banking institutions. (The remainder of these bonds are issued by
individual banks within BHCs.) The primary component of BHC supervision
is on-site inspections, which generally are conducted once a year, although
the largest and most complex BHCs have a continuous supervisory presence.
At the conclusion of a BHC inspection, the supervisors assign a composite
BOPEC rating that summarizes the five key areas of supervisory concern:
the condition of the BHC's Bank subsidiaries, Other nonbank subsidiaries,
Parent company, Earnings, and Capital adequacy. BOPEC ratings are assigned
according to an absolute scale from the highest rating of one to the
lowest rating of five; these ratings are confidential and are not made
public.
Can BHC debt spreads anticipate BOPEC changes?
If BHC debt market information
is to be useful to supervisors, it should agree with supervisory assessments
a reasonably large fraction of the
time. That is, large changes in BHC debt yields should give supervisors
an early warning of changes in BHC conditions. To examine this possibility,
we analyzed changes in BHC debt yields leading up to BOPEC assignments.
The yields were adjusted to account for multiple bond issues by the same
BHC, bond maturities, public credit ratings, and prior BOPEC ratings;
see Krainer and Lopez (2003). One caveat to our analysis is that bonds
are largely traded over-the-counter with greatly varying degrees of liquidity,
which should have an impact on the quality and reliability of the observed
bond yields.
 As shown in Figure 1, changes in adjusted BHC yields moved
in accordance with and up to a year prior to the future BOPEC assignments.
For the
BOPEC downgrades in our sample, the average cumulative increase in yields
was statistically significant at about one percentage point by the time
of the inspection, which is a signal that debt market investors were
demanding an increase in their investment return. For the sample's upgrades,
adjusted yields dropped by 75 basis points by the time of the inspection.
For the inspections without BOPEC changes, the yield changes were effectively
zero on average. We thus conclude that changes in adjusted BHC debt yields
are consistent with future supervisory assessments and could be useful
for supervisory monitoring.
Debt market information in an off-site monitoring
model
We gauged the potential usefulness of BHC debt market information
for supervisory monitoring by combining it with supervisory variables
in
an off-site monitoring model. Securities market data are, in general,
available sooner than supervisory data from quarterly financial statements,
which could assist an off-site monitoring model in detecting sudden changes
in BHC conditions. Since the cost of incorporating securities market
variables into the model is low, even small net improvements in forecast
accuracy could be valuable. We used our proposed BOPEC off-site monitoring
(BOM) model (Krainer and Lopez 2001). The benchmark, or core, BOM model
examines the relationship between BOPEC ratings and selected supervisory
variables; this model was extended to incorporate our adjusted BHC debt
yields. When the model was estimated over our full sample of BOPEC ratings
assigned from 1990 to mid-1998, adjusted BHC debt yields were statistically
significant and contributed to the model's empirical fit of the data.
BHC
debt yields, of course, represent only one source of market information.
As shown in Krainer and Lopez (2001), changes in BHC equity prices also
contain information useful for supervisory monitoring within the context
of the BOM model. When we extended the core model to include equity market
variables, both sets of securities market variables were statistically
significant and improved the model's empirical fit of the data more so
than models using just one set of market variables. This result suggests
that the BHC assessments expressed by both sets of investors via changes
in BHC securities prices could contribute to supervisory monitoring.
We
further examined whether securities market variables might provide different
information depending on how close a BHC is to default, measured
as the point at which BHC liabilities exceed assets. Using a measure
of a firm's default probability, commonly known as the distance-to-default,
based on the market value of BHC assets and outstanding liabilities,
we ranked the publicly traded BHCs in our sample and used those rankings
within the BOM model. We found that measures of equity returns had a
greater impact when BHCs were further away from default and that debt
yields had more impact when BHCs were closer to (but not in) default.
The asymmetric contributions of debt and equity return variables are
consistent with the theory that debtholders do not benefit directly from
gains accruing beyond the face value of their investment, but are affected
by actions that the BHC takes while relatively closer to default. In
contrast, equity holders should be less sensitive to changes in firm
value close to default because their investment might disappear entirely,
while further from default, they would be more sensitive to changes in
BHC condition.
To be useful for supervisory monitoring, the extended BOM
model with both BHC debt and equity market variables also must forecast
BOPEC ratings
accurately. To mimic actual practices, we re-estimated the BOM model
with and without securities market variables every quarter based on a
rolling four quarter sample of data. The estimated models were then used
to generate one-quarter-ahead BOPEC forecasts. They generated a forecasted
change in supervisory rating if the forecast was more than three-quarters
of a rating grade different from its corresponding lagged BOPEC rating.
When compared to all our sample's ratings at four quarters prior to assignment,
this extended model correctly signaled 67% of all the BOPEC assignments
and almost 60% of all BOPEC changes. These percentages increase to 70%
and 88%, respectively, at one quarter prior to assignment.
Another dimension
of accuracy for an off-site monitoring model is the mix of correct and
incorrect forecast signals; for example, if the model
signals a downgrade, what is the probability that the signal is correct?
This dimension of accuracy is measured using the ratio of correctly signaled
downgrades to the total number of signaled downgrades. We find "no
change" signals were the most common and were correct 67% of the
time. Downgrade signals, which are of more interest to supervisors, were
correct 58% of the time at four quarters prior and improved to 86% at
one quarter prior. Upgrade signals occurred with the same frequencies.
Our results indicate that forecast signals from the extended BOM model
were accurate a large percentage of the time, even up to a year prior
to the BOPEC assignments, and could thus be useful for off-site BHC monitoring.
A
critical question is whether forecasts from the extended model including
both sets of securities market variables provides further information
about BOPEC ratings beyond that obtained using forecasts from the core
model using only supervisory variables. To make this assessment, we
compared the accuracy of the two sets of forecasts statistically, which
in this
case is the percentage of BOPEC ratings accurately forecast. By this
measure, we found little difference between the accuracy of these two
sets of competing forecasts.
This result, however, does not mean that
the forecasted BOPEC ratings from the two models are identical. Since
the forecasting literature
has shown that combining forecasts from different models can improve
forecast
accuracy, we decided to gauge the contribution of securities market
information by examining the additional forecast signals for BHCs
with public securities
generated by the extended model relative to the core model's signals.
At four quarters prior, the extended model signaled 54 additional
BOPEC changes, of which 27 were correct and were almost evenly split
between
upgrades and downgrades. For one quarter prior, 65 additional BOPEC
changes were signaled, of which 42 (or 65%) were correct and again
almost evenly
split between upgrades and downgrades.
Seen in this light, the marginal
benefit of incorporating securities market variables into the BOM model
is notable. At four quarters
prior, the additional 27 correct signals regarding BOPEC changes
increased
the total to 93, a 40% increase. At one quarter prior, the additional
42
correct signals increased the total of correct BOPEC change forecasts
by over 30% to 175. The benefits from having the additional correct
early-warning signals provided by these forecasts could very well
be worth the supervisory
costs of dealing with the additional incorrect signals.
John Krainer
Economist |
Jose A. Lopez
Senior Economist |
References
[URLs accessed November 2003.]
Board of Governors of the Federal Reserve System and the U.S. Department
of the Treasury. 2000. "The Feasibility and Desirability of Mandatory
Subordinated Debt." Report to Congress pursuant to section 108 of
the Gramm-Leach-Bliley Act of 1999.
Krainer, J., and J.A. Lopez. 2001. "Incorporating Equity Market
Information into Supervisory Monitoring Models." FRBSF Working Paper
01-14.
Krainer, J., and J.A. Lopez. 2003. "How Might Financial Market Information
Be Used for Supervisory Purposes?" FRBSF Economic Review, pp. 29-45.
|
